SenseTime open-sources lightweight multimodal AI model SenseNova U1.5 Lite

SenseTime has open-sourced its lightweight, native unified multimodal model SenseNova U1.5 Lite. It supports 3-4k context length and handles multiple constraints like subject, quantity, and spatial relations to improve complex visual tasks.

The model boosts visual generation, image editing, text and layout handling, and 4K high-resolution output. It improves composition, color, and texture details, keeps subject identity and spatial structure, and enhances Chinese and English text layout.

As an 8B-parameter lightweight model, it outperforms similar-sized models in instruction following and image editing consistency, and approaches large-scale commercial models in text rendering and complex layouts. The model is open-sourced on GitHub, Hugging Face, and ModelScope.